Distilling Knowledge via Intermediate Classifiers
This addresses a specific bottleneck in knowledge distillation for machine learning practitioners, but it is incremental as it builds on existing distillation techniques.
The paper tackles the problem of knowledge distillation weakening when there is a large capacity gap between teacher and student models, by introducing intermediate classifier heads to create a cohort of teachers that co-teach the student, resulting in improved performance over canonical methods as shown in experiments on various datasets and model pairs.
The crux of knowledge distillation is to effectively train a resource-limited student model with the guide of a pre-trained larger teacher model. However, when there is a large difference between the model complexities of teacher and student (i.e., capacity gap), knowledge distillation loses its strength in transferring knowledge from the teacher to the student, thus training a weaker student. To mitigate the impact of the capacity gap, we introduce knowledge distillation via intermediate heads. By extending the intermediate layers of the teacher (at various depths) with classifier heads, we cheaply acquire a cohort of heterogeneous pre-trained teachers. The intermediate classifier heads can all together be efficiently learned while freezing the backbone of the pre-trained teacher. The cohort of teachers (including the original teacher) co-teach the student simultaneously. Our experiments on various teacher-student pairs and datasets have demonstrated that the proposed approach outperforms the canonical knowledge distillation approach and its extensions.